AX-EmbeddingGemma-2-MLX-AXQ-MXFP4 — 12.61 BPW measured main

An AXQuant (AXQ) mixed-precision MLX checkpoint for Apple Silicon, converted directly from the BF16 source model. The language path is quantized while the vision tower and audio tower are preserved at BF16 in the checkpoint (or a bound sidecar when present).

Development evidence — not a certified AXQuant release. This package has conversion and artifact-integrity records, but it does not publish measured quality, long-context, kernel-speed, or MTP-speed evidence. Do not interpret the AXQ product label as a benchmark claim.

Model details

Property	Value
Base model	google/embeddinggemma-2
Source revision	914f7f89142e33e77833254d9c9b90c3cef7303b
Product family	embedding-gemma-2
Source architecture	EmbeddingGemma2Model (dense); text path optimized
Main-model parameters	744.37M logical parameters
Quantizer	AXQuant 2.1.0
Hub budget class	MXFP4
AXQuant base precision class	16p0bpw
Planned storage-adjusted BPW	12.7220
Measured main-model BPW	12.6059
Measured total BPW	12.6059
Safetensors weight size	1.17 GB
Approximate complete download	1.21 GB
Configured maximum context	262,144 tokens; practical limits depend on unified memory
Primary MLX runtime	embedding runtime
AX Engine native execution	Not established; no validated native manifest is included
MTP present	False
Vision present	True
Audio present	True

This repository contains MLX Safetensors. It does not contain PyTorch or GGUF weights.

Choosing an AXQ pack

AXQ names describe a storage-budget product class, not one uniform precision applied to every tensor. Protected tensors remain at higher precision, so the exact measured BPW is authoritative. In particular, a 6bit-named mixed plan may retain 4bit as its base precision while selecting 6-bit, 8-bit, or BF16 for other tensors to meet an approximately 6-BPW total budget. Protection floors can also raise a 4bit-named pack close to (or above) a 6bit budget on small or heavily protected models. When that collapse happens, AutomatosX does not publish a separate misleading 4bit sibling for that base.

Sibling	Intended trade-off
MXFP8 sibling	Same AXQ role policy at the other MLX block format. Vision and audio stay BF16.

See the AutomatosX collections for the family catalog, or the complete index.

Download

python -m pip install -U huggingface_hub
hf download AutomatosX/AX-EmbeddingGemma-2-MLX-AXQ-MXFP4 --local-dir ./AX-EmbeddingGemma-2-MLX-AXQ-MXFP4

Allow at least 1.21 GB of free disk space. Pin the resulting Hub commit in reproducible deployments rather than relying indefinitely on main.

Run as an embedding model

This checkpoint returns embeddings. mlx_lm.generate and text-generation scores do not apply. Load it with the embedding runtime that matches the source family. Keep activations in BF16 or FP32. Rebuild retrieval indexes with this checkpoint; vectors from another precision are not interchangeable. The artifact records MLX 0.32.3.

Text, image, video, and audio share one 768-d embedding. Supported Matryoshka widths are 768, 512, 256, and 128; re-normalize after truncation. Use the task prefixes in config_sentence_transformers.json. Do not cast this model to float16. Load with an mlx-vlm build that provides embedding_gemma2 (mlx_vlm.embedding_loader.load_embedding_model). This development pack is not an MTEB result.

AX Engine status

This package does not include a validated native model-manifest.json, so AX Engine execution is not established by this release. The AX Engine fields in axquant_runtime.json describe the intended compatibility contract, not observed runtime evidence. Use the architecture-specific MLX runtime path above. The artifact records AX Engine version 7.6.3, but version discovery alone is not a runtime check.

Quantization layout

Main-weight precision	Parameters	Share
4bit	117.44M	15.78%
8bit	153.09M	20.57%
bf16	473.84M	63.66%

• Quantization methods: affine, bf16, mxfp4.
• Group sizes used by quantized assignments: 32, 64.
• MTP sidecar: not included.
• Vision sidecar: not included.
• Vision weights: protected BF16 in main shards.
• Audio weights: protected BF16 in main shards.
• Optimization scope: text-path.
• Support tier: convertible.

BF16 sidecars, when present, are included in total download size. Their presence does not by itself establish MTP acceleration or vision-language quality.

Evidence and validation status

Check	Status
Planning evidence	architecture_prior
Calibration	none; the allocation is based on architecture priors
Quantizer execution	218/218 recorded module conversions succeeded; 0 fallbacks
AX Engine native manifest	not included
Quality versus BF16 or uniform baselines	Not published; no quality-retention claim
MTP acceptance and speed	not measured; no MTP speedup claim
AX Engine kernel evidence	unmeasured
Vision-language quality	Not evaluated or claimed; vision tensors are preserved at BF16
Audio tower	Not evaluated or claimed; audio tensors are preserved at BF16
Long-context quality	262,144-token capacity is config metadata, not a validated claim
Release certification	Not certified; formal AXQuant M0-M8 gates are not closed

Intended use and limitations

• Intended for local development and evaluation on Apple Silicon with MLX-compatible runtimes.

• No minimum unified-memory figure is claimed; loadability depends on model size, context length, KV-cache policy, runtime buffers, and other processes using unified memory.

• Architecture-prior allocation is not measured sensitivity. It must not be presented as measured model quality.

• Vision weights are preserved at BF16, but this release does not claim validated VLM quality.

• Audio weights are preserved at BF16. This release does not claim retrieval quality for text, image, or audio inputs.

• The configured context window can require substantially more memory as the KV cache grows.

• AX Engine execution is not established because this package has no validated native manifest.

• Upstream capabilities, limitations, biases, and responsible-use guidance still apply.

Provenance and audit files

• axquant_manifest.json: package identity, byte accounting, runtime contract, software versions, and file checksums.
• axquant_plan.json: per-tensor precision decisions and planning evidence.
• axquant_quantizer_execution.json: conversion coverage and fallback records.
• axquant_runtime.json: declared AX Engine and MLX compatibility metadata; runtime checks remain separate evidence.

All published provenance uses repository-relative paths. Local source paths are stripped before publication. The checkpoint was converted from BF16 rather than re-quantized from an OptiQ artifact. If an OptiQ repository is published separately, it uses a different quantizer and should not be assumed to have identical BPW or quality.

License

The checkpoint follows the upstream model license where applicable (often Apache License 2.0). See LICENSE and the google/embeddinggemma-2 model card for license terms, model limitations, and responsible-use guidance.